Human-machine interaction

Abstract

A method for human-machine interaction based on a neural network is provided. The method includes: providing a user input as a first input for a neural network system; providing the user input to a conversation control system different from the neural network system; processing the user input by the conversation control system based on information relevant to the user input; providing a processing result of the conversation control system as second input for the neural network system; and generating, by the neural network system, a reply to the user input based on the first and second input.

What is claimed is: 
     
         1 . A method, comprising:
 providing a user input as a first input for a neural network system;   providing the user input to a conversation control system different from the neural network system;   processing the user input by the conversation control system based on information relevant to the user input;   providing a processing result of the conversation control system as a second input for the neural network system; and   generating, by the neural network system, a reply to the user input based on the first and second input.   
     
     
         2 . The method of  claim 1 , wherein the information relevant to the user input comprises work memory information that is valid only during current human-machine interaction and long-term memory information. 
     
     
         3 . The method of  claim 2 , wherein the long-term memory information comprises knowledge information in a form of a first directed graph comprising nodes and one or more directed edges, the nodes in the first directed graph are structured data comprising semantic content and logical control information, and each of the one or more directed edges in the first directed graph represents a relevance attribute between relevant nodes. 
     
     
         4 . The method of  claim 3 , wherein the logical control information comprises information for screening nodes relevant to the current human-machine interaction and/or information for determining the degree of relevance between the nodes in the current human-machine interaction. 
     
     
         5 . The method of  claim 3 , wherein each of the nodes in the first directed graph comprises a first type of nodes and a second type of nodes, the semantic content of the second type of nodes is a part of the semantic content of the first type of nodes relevant to the second type of nodes, and the logical control information of the second type of nodes comprises at least one selected from a group consisting of: the popularity of the second type of nodes under the first type of nodes relevant to the second type of nodes, a relevance skip relationship between the second type of nodes and at least one of other second type of nodes, and a subtype of the second type of nodes. 
     
     
         6 . The method of  claim 5 , wherein each of the nodes in the first directed graph comprises a third type of nodes, the semantic content of the third type of nodes supports multi-mode content, and the logical control information of the third type of nodes comprises information of the second type of nodes relevant to the third type of nodes and/or information for representing the semantic content of the third type of nodes. 
     
     
         7 . The method of  claim 3 , wherein the long-term memory information comprises conversation library information in the form of a second directed graph including nodes and one or more directed edges, and the second directed graph is isomorphic to the first directed graph. 
     
     
         8 . The method of  claim 3 , wherein the work memory information comprises information in a form of a third directed graph including nodes and one or more directed edges, wherein the third directed graph is isomorphic to the first directed graph and is a part of the first directed graph. 
     
     
         9 . The method of  claim 8 , wherein the work memory information comprises one selected from a group consisting of: semantic content and logical control information of all nodes relevant to the current human-machine interaction and taken from the first directed graph, and historical data of interaction record during the current human-machine interaction. 
     
     
         10 . The method of  claim 8 , wherein the work memory information comprises first information for marking a semantic content that has been involved in the current human-machine interaction. 
     
     
         11 . The method of  claim 10 , wherein the work memory information comprises second information for indicating a conversation party who first mentioned the semantic content that has been involved. 
     
     
         12 . The method of  claim 2 , wherein the processing result comprises a plan for replying to the user input in the current human-machine interaction situation. 
     
     
         13 . The method of  claim 12 , wherein processing the user input comprises:
 analyzing the semantic content of the user input; and   analyzing a communicative intention of the user corresponding to the user input in the current human-machine interaction.   
     
     
         14 . The method of  claim 13 , wherein analyzing the semantic content of the user input comprises:
 determining whether the user input is able to correspond to a certain node in the work memory information; and   in response to the user input is able to correspond to the certain node in the work memory information, processing the user input based on the work memory information.   
     
     
         15 . The method of  claim 14 , wherein processing the user input comprises:
 based on information of the certain node in the work memory information, supplementing relevant content for the user input.   
     
     
         16 . The method of  claim 14 , wherein analyzing the semantic content of the user input further comprises:
 in response to the user input is unable to correspond to the node in the work memory information, extracting information of a node relevant to the user input from the long-term memory information and storing the information in the work memory information.   
     
     
         17 . The method of  claim 13 , wherein analyzing the semantic content of the user input comprises:
 disambiguating the user input.   
     
     
         18 . The method of  claim 17 , wherein disambiguating the user input comprises:
 based on node information relevant to the current human-machine interaction in the user input and the work memory information, identifying at least part of content with ambiguity in the user input and determining the meaning of the at least part of content in the current human-machine interaction situation.   
     
     
         19 . The method of  claim 16 , wherein processing the user input further comprises:
 according to the semantic content of the user input and the communicative intention corresponding to the user input in the current human-machine interaction, querying information of the relevant nodes of the user input from the work memory information; and   according to the degree of relevance with the user input, sorting the relevant nodes of the user input acquired by query, wherein the sorting is performed based on the logical control information of the relevant nodes.   
     
     
         20 . The method of  claim 19 , wherein processing the user input further comprises:
 according to the degree of relevance with the user input, assigning different values to the relevant nodes.   
     
     
         21 . The method of  claim 19 , wherein the plan for replying to the user input in the current human-machine interaction situation comprises:
 according to the sorting result, planning a conversation target and selecting node information with the highest degree of relevance with the user input as a conversation content of the plan; and   integrating the conversation content of the plan and the conversation target as the second input and providing the second input for the neural network system.   
     
     
         22 . The method of  claim 21 , wherein the plan for replying to the user input further comprises:
 in the case where the work memory information is not updated for the user input and in response to the node with the highest degree of relevance is unable to meet a predetermined standard, re-querying the long-term memory information to update the work memory information.   
     
     
         23 . The method of  claim 1 , wherein the neural network system is an end-to-end neural network system, and wherein the end-to-end neural system comprises an encoder and a decoder, the encoder is configured to receive the user input and the stored historical interaction information of the current human-machine interaction, and the decoder is configured to receive the second input and generate the reply to the user input. 
     
     
         24 . An electronic device, comprising:
 one or more processors; and   a non-transitory memory storing one or more programs configured to be executed by the one or more processors, the one or more programs comprising instructions for:
 providing a user input as a first input for a neural network system; 
 providing the user input to a conversation control system different from the neural network system; 
 processing the user input by the conversation control system based on information relevant to the user input; 
 providing a processing result of the conversation control system as a second input for the neural network system; and 
 generating, by the neural network system, a reply to the user input based on the first and second input. 
   
     
     
         25 . A non-transitory computer-readable storage medium storing one or more programs, the one or more programs comprising instructions, which, when executed by one or more processors of an electronic device, cause the electronic device to:
 provide a user input as a first input for a neural network system;   provide the user input to a conversation control system different from the neural network system;   process the user input by the conversation control system based on information relevant to the user input;   provide a processing result of the conversation control system as a second input for the neural network system; and   generate, by the neural network system, a reply to the user input based on the first and second input.